When to open: You want to know how capital flowed in the AI Agent sector in 2025, what valuation levels looked like, who was placing bets, and who was warning of a bubble. This wiki is compiled by newtype as a complete 2025 market data report(data as of 2026-01-14).
Difference from AI Agent Technology Architecture Evolution 2025 (In-Depth Research Report) the difference: That report covers technology architecture,this wiki covers how the money flows. Reading both together gives a complete picture of the industry.
I. Market Size Overview
1.1 AI Agent Market Size Forecast
| Institution | 2025 Status | 2030/2033 Forecast | CAGR |
|---|---|---|---|
| MarketsandMarkets | $5.26 billion | $526.2 billion (2030) | 43.84% |
| Grand View Research | $7.63 billion | $182.97 billion (2033) | 49.6% |
"The AI agents market size was estimated at USD 7.63 billion in 2025 and is projected to reach USD 182.97 billion by 2033." — Grand View Research
1.2 Total Enterprise AI Spending (2025)
- Total investment: $37 billion (Menlo Ventures)
- Application layer: $19 billion (6% of the software market)
- Infrastructure: $18 billion
II. VC Investment Data (Full Year 2025)
2.1 Overall Scale
| Metric | Data | Source |
|---|---|---|
| Global GenAI VC Investment | $87 billion | EY Global Report |
| AI Agent-Specific Financing | $6.4 billion / 451 deals(as of October) | PitchBook |
| Year-over-year growth (vs 2024) | 65% | PitchBook |
| Average deal size | $49.3 million (2025) | 86% year-over-year growth |
| Mega-round share | Top 4 account for 30%+ of total | PitchBook |
| Share of $500 million+ rounds | 58% | PitchBook |
2.2 Top 10 Leading VC Firms
- Index Ventures
- Andreessen Horowitz (a16z) — Raised $15 billion in new funds (18% of total US VC funding)
- Sequoia Capital — Agent Economy thesis (three-stage theory)
- Lightspeed Venture Partners — Led Anthropic Series F
- Accel
- General Catalyst
- Kleiner Perkins
- IVP
- Founders Fund
- Tiger Global
2.3 a16z's Bets
- Raised $15 billion in new funds in 2025 (18% of total US VC fundraising)
- Led Thinking Machines Lab ($2 billion seed round)
- Investment focus: AI infrastructure, Agent application layer, Voice AI
"needed to ensure America wins the next 100 years of technology." — Ben Horowitz
2.4 Sequoia's Three-Stage "Agent Economy" Thesis
- Stage 1:AI Assistants(Assistants) — Current stage
- Stage 2:Agent Clusters(Agent Swarms) — 2026-2028
- Stage 3:Agent Economy(Agent Economy) — 2028+
Representative investments:
- Harvey (legal AI): $300 million Series D, $3 billion valuation
- Serval (IT Agent): $1 billion valuation
Strategist: Konstantine Buhler
III. "Kingmaking Strategy"
The most anomalous phenomenon in 2025: startups completing two consecutive funding rounds within 27–60 days, with valuations doubling or even tripling without new data to support the increase.
3.1 Typical Case: Cursor
| Time | Valuation | Notes |
|---|---|---|
| January | (Undisclosed) | |
| June | $9.9 billion | $900 million financing |
| November | $29.3 billion | $2.3 billion Series D (3x growth in 5 months) |
| Annualized revenue | $1 billion | Sales revenue grew 100x |
3.2 Lovable(Vibe Coding)
| Time | Valuation | Notes |
|---|---|---|
| July | $1.8 billion | $200 million Series A (8 months after founding) |
| December | $7 billion | $330 million financing (nearly 4x growth in 4 months) |
3.3 OpenEvidence (Medical AI)
| Time | Valuation |
|---|---|
| February | $925 million |
| July | $3.5 billion ($210 million Series B) |
| October | $6 billion(6.5x growth in 8 months) |
"There's no new data between rounds. Series Bs happen 27-60 days after Series As regularly." — Jaya Gupta, Foundation Capital
IV. Foundation Model Giant Financing Race
4.1 OpenAI — The Front-Runner
| Time | Valuation | Notes |
|---|---|---|
| 2025-03 | $300 billion | $40 billion financing (led by SoftBank) |
| 2025-09 | $500 billion | $6.6 billion secondary market |
| 2025-12 (rumored) | $830 billion | $100 billion financing in talks |
| 2026 IPO target | $1,000 billion |
- Cumulative financing: $66.4 billion (as of 2025-08)
- 2025 revenue target: $13 billion
- Actual run rate:$20 billion(50% above expectations)
- Barclays forecasts 2027 ARR:$100 billion
4.2 Anthropic — Fast Follower
| Time | Valuation |
|---|---|
| 2025-03 | $61.5 billion (Series E, $3.5 billion) |
| 2025-09 | $183 billion(Series F, $13 billion, led by Iconiq) |
| 2026-01 (latest) | $350 billion($10 billion new round, led by GIC + Coatue) |
- Cumulative financing: $39.1 billion
- Revenue: $1 billion at start of 2025 → $5 billion in August$5 billion(5x growth in 8 months)
- IPO preparation: Has retained Wilson Sonsini
4.3 xAI (Elon Musk) — Rising Star
- Time since founding: half
- Cumulative financing:$42.1 billion(has surpassed Anthropic)
- 2026-01 Series E:$20 billion(exceeded the $15 billion target)
- Lead investors: Valor Equity, StepStone, Fidelity, QIA
- Strategic investors: Nvidia, Cisco
- Valuation: Undisclosed, rumored at $200 billion
4.4 Other Foundation Models
| Company | Founder | Data |
|---|---|---|
| Safe Superintelligence | Ilya Sutskever | $2 billion seed, $32 billion valuation,20 people |
| Thinking Machines Lab | Mira Murati (former OpenAI CTO) | $2 billion seed (largest in history), $12 billion valuation |
| Mistral AI(France) | — | $1.5 billion |
V. Sub-Sector Financing
5.1 Coding / Development Tools
| Company | Valuation | ARR | Revenue Multiple |
|---|---|---|---|
| Cursor | $29.3 billion | $1 billion | 29x |
| Cognition(Devin AI) | $10.2 billion | $150 million+ | 68x |
Cognition's per-employee valuation: $98.1 million
5.2 Healthcare
| Company | Data |
|---|---|
| Abridge(doctor-patient transcription) | $250 million Series D, $2.75 billion valuation (led by IVP) |
| Hippocratic AI | $126 million Series C |
5.3 Legal Tech
| Company | Data |
|---|---|
| Harvey | $300 million Series D, $3 billion valuation (led by Sequoia) |
| Eudia | $105 million Series A (led by General Catalyst) |
5.4 Enterprise Software
| Company | Data |
|---|---|
| Together AI(open-source GenAI) | $305 million Series B, $3.3 billion valuation |
| Lambda(AI infrastructure) | $480 million Series D, $2.5 billion valuation (Nvidia participated) |
| Mercor(AI recruitment) | $450 million (two rounds), $10 billion valuation |
5.5 Prediction Markets
| Company | Data |
|---|---|
| Kalshi | June $2 billion → Series E $1 billion financing, valuation $11 billion(5.5x growth in 6 months) |
VI. Unicorn Birth Wave
6.1 Numbers
- Total new unicorns in 2025:191 companies(globally)
- 2024: 128 companies
- Total AI unicorns (as of end of 2025):498 companies
- Total valuation:$2.7 trillion
6.2 The "Ultra-Small Team Unicorn" Phenomenon
10 AI companies with valuations of $1 billion+ and teams of ≤ 50 people:
| Company | Valuation | Headcount |
|---|---|---|
| Safe Superintelligence | $32 billion | 20 |
| Magic | $1.58 billion | 20 |
| 0G Labs | $2 billion | 40 |
| Lovable, Mercor,... | — | — |
"We're going to see 10-person companies with billion-dollar valuations pretty soon... there is a one-person billion-dollar company, which would've been unimaginable without AI." — Sam Altman
6.3 New European Unicorns (2025)
- Lovable(Sweden): $1.8 billion
- Isomorphic Labs(UK): $600 million financing (Google DeepMind spin-off)
- Tines(Ireland): $1.125 billion
- Tekever(Portugal): £1 billion+
- Quantum Systems(Germany)
VII. Rise of Sovereign Wealth Funds
7.1 Investment Scale
- 2025 sovereign fund AI investment:$46 billion
- Motivation: Economic diversification + technological autonomy ("Sovereign AI")
7.2 Major Players
| Sovereign Fund | Investment Target |
|---|---|
| Qatar Investment Authority | Anthropic, xAI |
| Singapore GIC | Led Anthropic's $10 billion round |
| Saudi Project Transcendence | $100 billion commitment |
7.3 Government Funding Commitments (2024–2025)
| Country | Data |
|---|---|
| Saudi Arabia | Project Transcendence $100 billion |
| China | Semiconductor fund $47.5 billion |
| France | AI plan €109 billion |
| Canada | $2.4 billion |
| India | $1.25 billion |
VIII. Analyst Views
8.1 Gartner
Enterprise Agent Adoption Rate:
| Year | Adoption Rate |
|---|---|
| 2025 | < 5% |
| 2026 | 40%(task-specific Agents) |
| 2028 | 80% (AI Copilot embedded) |
Revenue Forecast:
- 2025: Agentic AI accounts for 2% of enterprise application software revenue
- 2035:30%(over $450 billion)
Sales Domain:
- 2028: Number of Agents = 10x the number of human salespeople10 times
- But only <40% of salespeople believe Agents improve productivity
8.2 McKinsey
Economic Impact: GenAI contributes $2.6–4.4 trillion to annual global GDP$2.6–4.4 trillion
Enterprise Adoption Status (2025):
- 88% of organizations use AI in at least one function (2024: 78%)
- 23% enterprises have deployed Agents "at scale" somewhere
- 39% are experimenting
- Only high-performing enterprises 6%(EBIT impact ≥ 5%)
Agent Value Share:
- 2025: Agents contribute of total AI value17%
- 2028 forecast:29%
8.3 BCG
Productivity Improvement:
- Low-value work time reduced: 25–40%
- Business process acceleration:30-50%
Value Gap:
- Only 5% companies realize full AI value
- 60% of enterprises see almost no substantive return
- "Future-built" companies: 15% of AI budget allocated to Agents
Customer Case ROI:
| Industry | Improvement |
|---|---|
| Global banks | AI customer service reduces costs 10x |
| Pharmaceuticals | R&D cycles shortened by 25%, efficiency improved by 35% |
| IT departments | Productivity improved 40% |
8.4 IDC
- 2026:80% enterprise workplace applications embed AI Copilots
8.5 CB Insights(2025 Q3)
- AI Agent market leads deal volume and M&A activity
- 3 of the top 5 AI exits are Agent-related:
- Workday acquires Sana Labs
- NICE acquires Cognigy
- Atlassian acquires The Browser Company (Arc/Dia)
IX. Application Layer vs Infrastructure Layer (Menlo Data)
9.1 Application Layer Breakdown ($19 billion)
Horizontal AI: $8.4 billion:
- Copilots:$7.2 billion (86%) — ChatGPT Enterprise, Claude for Work, Microsoft Copilot
- Agent platforms: $750 million (10%) — Salesforce Agentforce, Writer, Glean
- Personal productivity: $450 million (5%)
Vertical AI: $3.5 billion(3x that of 2024):
- Healthcare: leading
- Legal: $650 million
- Creator tools: $360 million
- Government: $350 million
9.2 Infrastructure Layer ($18 billion)
- Foundation model APIs: $12.5 billion
- Model training: $4 billion
- AI infrastructure: $1.5 billion
X. Valuation and Cash Burn Multiples
10.1 Valuation Multiples (Revenue Multiples)
| Company / Category | Multiple |
|---|---|
| Traditional SaaS | 5-10x |
| Early-stage AI companies | 10-50x |
| OpenAI ($500 billion valuation / $20 billion revenue) | 25x |
| Nvidia (benchmark) | 24x |
| Cognition / Devin($10.2 billion / $150 million) | 68x |
10.2 Burn Multiple
- Median Series A for AI companies:Burns $5 for every $1 of new revenue
- Far higher than other industries
"Burn multiples for AI companies are higher than other sectors, suggesting low-cost capital could be fueling inefficient growth." — SVB H2 2025
10.3 Deal count vs. single-deal size
- Deal count: down year-over-year 35%
- Average deal size: up 86%
- $100 million+ rounds: share of recent financing 75%+
Capital is concentrating toward "perceived winners"; the median early-stage deal rose from $2.5 million to $3.4 million
XI. Bubble Warnings
11.1 PitchBook warning
"Market value concentration indicates an increase in long-term systemic risk to venture capital... revenue multiples reach unsustainable levels."
11.2 Exit challenges
- Late-stage investors face enormous "exit barriers"
- Sky-high IPOs are required for late-stage investors to break even
- Contradiction between early enterprise adoption rates and sky-high valuations
11.3 DeepSeek shockwave (2025-01)
- China's DeepSeek developed a model matching OpenAI's at 1/20-1/40 cost
- Triggering VC reflection on the "moat thesis"
- Marc Andreessen called it a "profound gift to the world"
11.4 Disconnect between revenue and valuation
| Company | Contradiction |
|---|---|
| Cognition | ARR $150 million+, valuation $10.2 billion (68x) |
| Thinking Machines | Valued at $12 billion before product launch |
| Figure (humanoid robot) | Zero revenue, valued at $39 billion |
11.5 Investor divergence
- Bullish camp: "exponential event for vertical SaaS" (Garry Tan, Y Combinator)
- Bearish camp: "a lot of these companies marked as unicorns are not truly unicorns" (Michael Paulus, PCM Encore)
11.6 Geopolitical risks
"not thrilled about taking money from dictatorial Gulf states" — Dario Amodei (on QIA investment, internal memo)
XII. Sequoia's 12–18 Month Investment Thesis
5 key focus areas(Sequoia AI Investment Thesis 2025):
Persistent Memory
- Agent cross-session context retention
- Identity/personality persistence
Seamless AI-to-AI Communication protocols
- Inter-agent collaboration standards beyond MCP
- May erode the moats of "one-click solution" providers
AI Voice
- Real-time low-latency conversation
- Consumer-facing: AI companions, therapists
- Enterprise-facing: logistics coordination, trading desk automation
AI Security
- Full-stack security requirements
- Increased autonomy → expanded attack surface
Open Source AI
- Ensuring future decentralization
XIII. Key Quotes Compilation
13.1 VC perspectives
"If 2024 was the year of LLMs, we believe 2025 will be the year of agentic AI." — Praveen Akkiraju, Insight Partners
"Smart Money is all in on AI agents. Top 25 VCs backed 13 AI agent startups in September, representing nearly 20% of total early-stage activity." — CB Insights, 2025-09
13.2 Analyst perspectives
"By 2028, AI agents will outnumber human sellers by tenfold—yet fewer than 40% of sellers will report AI agents improved productivity." — Gartner, 2025-11
"Only 5% of companies are realizing AI's full value at scale. The widening gap." — BCG Build for the Future 2025
13.3 Industry leaders
"AI isn't a feature anymore. It's infrastructure." — McKinsey State of AI 2025
"If you're still piloting, you're dangerously behind." — Bain Technology Report 2025
XIV. Core Conclusions of the Report
2025 is a watershed year for AI Agent investment:
- Capital frenzy: $87 billion in GenAI investment hits a new high, with $6.4 billion dedicated to Agents
- Valuation surge: 191 new unicorns with a total valuation of $2.7 trillion; OpenAI sprinting toward $1 trillion
- Institutional bets: Top VCs allocate 15–20% of their portfolios to Agents; sovereign funds invest $46 billion
- Application explosion: Gartner predicts 40% of enterprise applications will embed Agents by 2026
- Value divergence: Only 5% (BCG) or 6% (McKinsey) of enterprises achieve scaled AI value
Shift in investment logic: From "what can AI do" → "who can Agents replace" — programming, law, healthcare, sales are the four hottest tracks.
Core contradiction: Sky-high valuations vs. low enterprise adoption + open-source challenges such as DeepSeek = 2026 may usher in a bubble test period.
Sequoia's judgment: The Agent Economy will create a trillion-dollar market by 2030 — this is the fundamental reason VCs are going all-in.
XV. Implications for independent developers / small studios
- Don't bet on "raising funding to compete with Cursor" — capital concentration is 30% in the Top 4; small studios cannot land mega rounds
- Vertical tracks have a window — 2025 vertical AI is 3x that of 2024, but market size is still small, so first-mover advantage is valuable
- Keep an eye on Sequoia's 5 areas — Persistent Memory, A2A, Voice, Security, Open Source; any of them can support a small-but-beautiful business
- Embrace the byproducts of the "king-making strategy" — Big companies have valuations pushed up by hype and large budgets to spend on small and mid-sized vendors
- The DeepSeek approach is worth learning from — Build matching products at 1/20 the cost; small studios can use the same logic to achieve local optima in narrow scenarios
- Low enterprise adoption = large sales opportunities — Only 23% have scaled, 77% haven't started; Agent SaaS for SMEs has room
History
- 2026-05-16: Phase 3 ingest from
12_AI_Agent_市场投资趋势_2025.md(created 2026-01-14). Full coverage of market size + VC / sovereign fund / unicorn data + foundation model valuation timeline + sub-tracks + Gartner/McKinsey/BCG/IDC perspectives + bubble warnings + Sequoia thesis + implications for independent developers.